Aon plc is a leading global firm providing risk, reinsurance, retirement, and health solutions. Focusing on data-driven insights, Aon operates in over 120 countries. For Aon Reinsurance Solutions, Boby Azarbod, the Data Services Lead, and Anthony Meyers, AI + Data Strategy Lead, spearhead efforts to modernize data infrastructure, enhancing operational efficiency and client value.
In the insurance sector, data enablement is vital for improving risk assessment, underwriting accuracy, and fraud detection. But data, in the abstract, can be unwieldy and messy. Improving data enablement, such as data visibility and data sharing, is vital to achieving a positive outcome. For instance, sharing claims data across business units can help build robust models that enhance pricing strategies and streamline claims processing. This collaborative approach allows businesses to access real-time data, reducing delays and improving ultimate decision-making.
As Anthony likes to say, “Data at rest is worthless.” So how do we make data active? It starts with data architectures, and the best architectures for a solution are often the simplest.
Aon is dedicated to establishing a sustainable data lakehouse strategy by aligning with Databricks' 6 Principles for Effective Data Lakehouse. Principle # 2 is “Remove Data Silos and Minimize Data Movement”. Data movement, copy, and duplication take time and may decrease the quality of the data in the Lakehouse, especially when it leads to data silos. Delta Sharing is a push in that direction.
Aon carefully evaluated the trade-offs between centralized and decentralized data strategies to optimize its Lakehouse architecture. Inspired by the recommendation of “Domain-oriented decentralized data ownership and architecture”, Aon adopted Delta Sharing as the backbone for seamless integration across decentralized Lakehouses, ensuring data sharing remains secure, efficient, and sustainable. This strategy minimizes the total cost of ownership per team while fostering collaboration and maintaining a unified governance framework.
The below highlights some improvements over our existing approaches that leverage an SFTP approach.
Aon has implemented Delta Sharing across various teams such as GRiDs, Risk Capital Analytics Platform, Impact Forecasting, and, many of its market intelligence platforms.
Aon uses Delta Sharing to significantly improve its risk modeling capabilities. By sharing real-time risk data across teams, Aon can integrate diverse datasets—such as climate models, economic indicators, and historical claims data—into its risk assessments. This integration allows for more precise predictions of potential losses and better-informed decisions regarding risk transfer strategies.
Delta Sharing enhances Aon's ability to build and maintain products by enabling real-time access to massive-scale data. This allows real-time risk analysis and prescriptive analytics by dramatically simplifying the I/O and cleaning practices. Delta Sharing also supports the change management process, creating extreme efficiencies if upstream changes need to be made.
Aon uses Delta Sharing to facilitate secure data exchange. This capability supports joint ventures by providing partners, internal and external, with timely access to relevant datasets without requiring them to be on the same platform. This approach not only enhances service offerings but also reduces operational costs associated with traditional data sharing methods
Aon's Delta Sharing initiative has resulted in three key business outcomes, each driven by the unique capabilities of Databricks Delta Sharing.
Here are three key learnings that Aon wants to share with the data sharing community
When implementing Delta Sharing, it’s essential to allocate time and resources to align with internal stakeholder teams, including legal and compliance. These teams will play a critical role in ensuring the solution is accepted and endorsed across the organization. Avoid setting rigid deliverables when dependencies on these teams are involved, as their input is crucial for long-term success.
If your organization uses multiple cloud platforms like AWS, GCP, and Azure, ensure that sharing works seamlessly across them. Proactively establish connections between them to enable smooth data sharing and collaboration, maximizing the value of your lakehouse architecture.
Beyond sharing raw data, we recommend sharing AI models across platforms to drive greater collaboration and innovation. Aon develops models in Databricks on AWS and consumes those models in Azure Databricks through Delta Sharing, eliminating the need for duplicating code repositories. This approach streamlines workflows and enhances cross-platform model deployment efficiency.
Aon plans to expand its use of Delta Sharing by exploring additional opportunities: